INDEX

Explanations

locations and associated entities

The neuron strongly activates on capitalized tokens that are parts of multi‐word proper names—especially place names, building names, streets, landmarks, and other named entities.

New Auto-Interp

Configuration

Prompts (Dashboard)

24,576 prompts, 128 tokens each

Dataset (Dashboard)

monology/pile-uncopyrighted

Embeds

IFrame

Link

Not in Any Lists

Negative Logits

 Encuentra

-1.18

HAW

-1.16

 Commencez

-1.16

VIK

-1.15

ではある

-1.12

antro

-1.12

gemä

-1.11

 İş

-1.10

事が

-1.10

of

-1.10

POSITIVE LOGITS

）：

1.51

<bos>

1.40

《

1.30

！）

1.28

érios

1.27

 that

1.26

！！！！

1.24

 seguras

1.23

*/

1.19

 adecuada

1.16

Activations Density 0.119%